Several kinds of charged particle beam exposure systems are known in the art. Most of these systems are provided to transfer very precise patterns onto an exposure surface of a substrate. Since lithography features are pushed to become smaller and smaller following Moore's law, the high resolution of electron beams could be used to continue the drive to even smaller features than today.
A conventional Gaussian charged particle beam exposure apparatus has a throughput of about 1/100 wafer/hr. However, for lithography purposes a commercially acceptable throughput of at least a few wafers/hr is necessary. Several ideas to increase the throughput of such an apparatus have been proposed.
U.S. Pat. Nos. 5,760,410 and 6,313,476 for instance, disclose a lithography system using an electron beam having a cross section, which is modified during the transferring of a pattern to an exposure surface of a target. The specific cross section or shape of the beam is established during operation by moving the emitted beam inside an aperture by using electrostatic deflection. The selected aperture partially blanks and thereby shapes the electron beam. The target exposure surface moves under the beam to refresh the surface. In this way a pattern is written. The throughput of this system is still limited.
In US-A1-20010028042 US-A1-20010028043 and US-A1-20010028044 an electron beam lithography system is disclosed using a plurality of electron beams by using a plurality of emitters or sources, each emitter provided for generating one electron beamlet. Each beamlet is then individually shaped and blanked to create a pattern on the underlying substrate. As all these emitters have slightly different emission characteristics, homogeneity of the beamlets is a problem. This was corrected by levelling every individual beam current to a reference current. Correction values for the mismatch are extremely difficult to calculate and it takes a significant amount of time, which reduces the throughput of the system. This especially turns problematic when using up to about 13.000 beamlets.
In GB-A1-2.340.991 , a multibeam particle lithography system is disclosed having a single source illumination system, which produces a plurality of charged particle sub-beams. The illumination system uses either a single ion source with aperture plates for splitting a beam in sub-beams, or a plurality of sources each producing a beam which is focused and projected. In general, the sources disclosed do not have a sufficient brightness.
In Jpn. J. Appl. Phys. Vol. 34 (1995) 6689-6695 a multi-electron beam (‘probes’) lithography system is disclosed having a single, specific ZrO/W-TFE thermal emission source with an emitter tip immersed in a magnetic field. The source has sufficient brightness, but a disadvantage of such a source is its limited total current. Furthermore, this source needs a crossover. The mutual homogeneity of the ‘probes’ is not further discussed. Furthermore, the intensity of the source is a not sufficient for about 13,000 beamlets.
In practise, many different approaches were proposed for multi-beam exposure systems. In one approach, one single source is used. The beam resulting from this source is split into many beamlets. In this approach, one single collimator lens is used. This approach has several drawbacks. For one, a large collimator lens suffers from aberrations. Furthermore, it proved difficult to obtain a single source which is at the same time very bright and has sufficient total emission current for a large number of beamlets.
In another approach, instead of the single collimator lens, the beam of a single source is split into a plurality of beamlets. Each beamlet is then individually deflected in such a way that substantially parallel beamlets are obtained. This will lead to large deflection angles for some of the beamlets and thus very difficult engineering challenges. Furthermore, again, it proved difficult to obtain a single source which is at the same time very bright and has sufficient total emission current for a large number of beamlets.
In yet another approach, n sources and n collimator lenses are used. A disadvantage of this approach is that is difficult to obtain the required number of beamlets per area of surface of the substrate: lenses for charged particle beams usually have a diameter which is about 10 times larger than the diameter of a beam. The different groups of beamlets will thus be spread over a large surface area.